• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

NVIDIA Could Tease Its Next-Gen Ampere GPU on 7nm at GTC 19

So basically Nvidia "could" have done a lot of things during GTC 2019 but instead decided to do "not a lot" for discreet GPU gaming hardware and just used it as a platform to market RTX hardware with a bit more messaging.

Or to put it another way they are going to publicly do naff all until AMD show Navi in action.

GTC isn't really geared towards consumer graphics, in fact they showed a lot more about that than anyone thought they would before the conference.

For gamers and the like, gotta wait for Computex and Gamescom.
 
Nvidia taped out Ampere last week, likely on TSMC N7+.

bPOpUqA.png


RAxXgeu.png



We could see Ampere GPU launch later this year, Turing was probably first taped out in mid 2017 on Samsung 10nm but Nvidia not very happy with yield then redesigned Turing to taped out on TSMC 12nm probably in very early 2018.
 
It is Odd, but it just highlights why you need dedicated hardware and justifies to RTX owners why they paid for the 20 series.
I expect it is related to the Crytex AMD RT demo and there will be mindless AMD fanboys proclaiing that yoy don;t need RTX GPUs, AMD are so amazing they can do RT without dedicated harwdare, despite the fact Nviodia already demosnatretd that last year with Volta.

What this release shows is that Nvidia can of course develop a software only RT solution but the whole point of RTX is it is 2-3x faster and actually gets you playable framerates.
Were you in a blind rage while typing about AMD fanboys?
 
I've actually been wanting to upgrade my system to a dual GPU system. The question is do I buy a second hand 1080Ti to go along with my existing 1080Ti or do I wait for a 2180Ti to come out and buy two of those?
 
Nvidia taped out Ampere last week, likely on TSMC N7+.

bPOpUqA.png


RAxXgeu.png



We could see Ampere GPU launch later this year, Turing was probably first taped out in mid 2017 on Samsung 10nm but Nvidia not very happy with yield then redesigned Turing to taped out on TSMC 12nm probably in very early 2018.

OK and so it begins.

Whoo whoo, all aboard the hype train.

Insert suitable gif in here( I can't be bothered because I'm on my phone :D:p:D)
 
Wonder how many will jump ship to the new cards when these come out, even though i remember a few saying they got their 2xxxx card for upcoming games like Cyberpunk.. Bet the next series of cards will be out by the time those games are out too!

Not touching any of the current cards, even if the next lot are still expensive they are bound to be much better IMO.
 
Nvidia taped out Ampere last week, likely on TSMC N7+.

bPOpUqA.png


RAxXgeu.png



We could see Ampere GPU launch later this year, Turing was probably first taped out in mid 2017 on Samsung 10nm but Nvidia not very happy with yield then redesigned Turing to taped out on TSMC 12nm probably in very early 2018.

Early April 1st?

Problem is that I can believe this :rolleyes:, but N7+?
 
Guys, from taping out to actual release it's going to be another year. Relax, nothing's coming any time soon. And it's not a gaming arch anyway, so unless you have $3-5k burning a whole in your pocket - meh.
 
Guys, from taping out to actual release it's going to be another year. Relax, nothing's coming any time soon. And it's not a gaming arch anyway, so unless you have $3-5k burning a whole in your pocket - meh.

Volta is great for gaming.:p

Having said that Turing is really bad at DP.:D

All joking aside you are right.:)
 
nVidia have been looking closely at TSMC 7nm as well as two different nodes at Samsung amongst other options - supply and unit cost are big considerations at the moment.


Nvidia have also publicly stated the current 7nm offerings are not interesting enough because of the huge costs but relatively small performance gains. The 7nm+ nodes should all actually be cheaper to produce with better yields and another 10-15% performance. Hence nvidia have stuck with 12nm.
 
Nvidia have also publicly stated the current 7nm offerings are not interesting enough because of the huge costs but relatively small performance gains. The 7nm+ nodes should all actually be cheaper to produce with better yields and another 10-15% performance. Hence nvidia have stuck with 12nm.

There was a very believable story about Nvidia having tried to place Turing on 10nm(37.5% improvement), which failed forcing them back to 12(25%). I'm not suggesting that node size directly equates to performance, but isn't the 2080Ti ~27% faster than the 1080Ti and a little behind the norm of what we have come to expect in an upgrade? ~35% would have been much more welcome.

Nvidia's marketing plays very well to the paper towel brigade, hence IMO the recent shots towards AMD on 7nm.
 
Well definitely another reason to avoid 2080Ti, it's like a proper lemon of a card I'm glad I keep resisting the urges to get a 2080.
 
Guys, from taping out to actual release it's going to be another year. Relax, nothing's coming any time soon. And it's not a gaming arch anyway, so unless you have $3-5k burning a whole in your pocket - meh.

No it not going to be 1 year from taping out to release if you looked at Nvidia past GPUs track record.

Kepler GK104 taped out in June 2011 to launch 22 March 2012 revision A2 294mm2 die TSMC 28nm took about 9 months with 2 revisions from taped out to launch. If GK104 A1 had no problems then it would take about 5 months from taped out to launch.
Maxwell GM204 taped out in April 2014 to launch 19 Sept 2014 revision A1 398mm2 die TSMC 28nm took about 5 months from taped out to launch.
Pascal GP104 taped out in Dec 2015 to launch 22 May 2016 revision A1 314mm2 die TSMC 16nm took about 6 months from taped out to launch.
TSMC 12nm went online for risk production first time in Nov 2016, Volta GV100 taped out around Nov 2016-Jan 2017 to launch 21 June 2017 revision A1 815mm2 die TSMC 12nm took about 6-8 months from taped out to launch.

There are no information on Turing tapeout between Samsung 10nm and TSMC 12nm is very difficult so I will have to guess the timeline.

https://www.techpowerup.com/247657/...s-in-2019-turing-originally-intended-for-10nm
https://news.samsung.com/global/sam...2nd-generation-10nm-finfet-process-technology

https://translate.google.com/translate?sl=auto&tl=en&u=https://pc.watch.impress.co.jp/docs/column/kaigai/1143/278/index.html

PC Watch interviewed Nvidia:

By the way, it is said that NVIDIA originally intended to use Samsung's 10 nm process in this generation. By staying at the mature 12 nm, it seems that we chose to increase the yield even if the die is enlarged.

Nvidia was probably originally planned to launched Turing TU104 first around Nov 2017 on Samsung 2nd gen 10nm so it would taped out in June 2017 with similar die size as Pascal but the yield was so bad and Nvidia not happy with it and probably took Nvidia a few months redesigned Turing TU104 to used TSMC 12nm and taped out in April 2018 and finally launched late 5 months later on 20 Sept 2018 alongside with TU102 RTX 2080 Ti.

I think Nvidia probably will not repeat the same mess with Turing this time with Ampere so Nvidia will launch small GA104 chip first then big chip GA102 10 months later.
 
Last edited:
Wonder how many will jump ship to the new cards when these come out, even though i remember a few saying they got their 2xxxx card for upcoming games like Cyberpunk.. Bet the next series of cards will be out by the time those games are out too!

Not touching any of the current cards, even if the next lot are still expensive they are bound to be much better IMO.

Since it will probably hammer the latest hardware in the same way Witcher 3 did at the time. I bet there will be a lot of "Cyberpunk 2077 killed my 2080" complaints. When they start being thrashed many more will start to fail.
 
16gb Gddr6 vram

Price depends on how much faster does Navi get?

Lord have mercy on us if Navi doesn’t get faster than 2070 level - I guess my 4K 120hz would be stuck at 60 to 80hz
 
Back
Top Bottom